The Reliability of the SASSI-3; A Summary

February 10, 2000

Sample
These findings are based on the results of 2,646 respondents. All of these participants had scored satisfactorily on a measure designed to identify individuals who did not respond to the SASSI-3 items in a meaningful manner (i.e., the "Random Answering Pattern" [RAP] scale of the SASSI-3). The majority of this sample (74%) consisted of clients from a variety of treatment programs throughout the country, including addictions treatment centers, a psychiatric hospital, a dual diagnosis hospital (substance abuse and psychological), a sex offender treatment program and a vocational rehabilitation program. A small percentage of these subjects (2%) were prisoners in a country detention center or members of a group who responded to newspaper ad requesting subjects with a family history of alcohol use. The remaining participants (24%) were from a number of non-treatment settings, such as colleges and nursing schools. Sixty-two percent of the total sample was male. Fifty-nine percent of the sample was Caucasian, 22% was African-American, and 11% was Native American. Forty-four percent reported never being married, whereas 29% were married and 23% divorced. The average age of this group was 32, and the average educational level was twelfth grade. Thirty-four percent of the sample (n = 888) had been interviewed and diagnosed by a trained clinician. Of those interviewed, 80% were diagnosed as having some type of substance use disorder.

Overall and Scale Reliability
Table 1 displays alpha coefficients for the SASSI-3 and its subscales. The coefficient alpha value for the SASSI-3 inventory was found to be .94 (see table 1). Test-retest coefficients obtained with a sample of 40 other respondents ranged from .92 to 1.00 (see table 2). The time between administrations of the test ranged from one to two weeks. It is important to recognize that the scales of the SASSI-3 were developed with the primary purpose of maximizing security. Items that contributed substantially to the accuracy of the SASSI-3 were incorporated into the questionnaire regardless of their impact on traditional measures of internal consistency (e.g., coefficient alpha) or on the dimensionality (i.e., the number of general characteristics assessed by a measure) of the scales. The coefficient alpha statistic (a statistic based on the average correlation among items and the number of items that is used to estimate the internal consistency of a unidimensional measure) is therefore of little importance for the scales of the SASSI-3 that were not designed to be unidimensional in nature.

Validity of Classifications on the SASSI-3
The results of the SASSI-3 were compared to the diagnoses obtained from clinicians. A correspondence rate (i.e., accuracy rate) of 93.8% was obtained (see table 3). The sensitivity rate of the SASSI-3 (i.e., the percentage of respondents diagnosed as having a substance-related disorder who were identified as such by the SASSI-3) was 94.1% (see table 3). The specificity rate (i.e., the percentage of respondents diagnosed as not having this type of disorder who were classified as such by the questionnaire) was 92.7% (see table 3).

To further evaluate the construct validity of the SASSI-3, the relationships of the inventory with a number of other criteria were examined. These analyses generally supported the premise that the SASSI-3 measures the likelihood of having a substance use disorder. For example, the SASSI-3 was found to be related to clinical diagnoses of substance abuse history. Ninety-nine percent of those testing positive on the SASSI-3 were evaluated by a clinician as having abused alcohol and/or one or more illicit drugs (although not necessarily currently having a substance-related disorder). In contrast, only 20% of those testing negative on the SASSI-3 were diagnosed as having abused some substance (chi-squared (1) = 185.53,p<.001). The SASSI-3 was also associated with self reports of alcohol and drug-related arrests (t = 3.96,p<.001), the number of illicit drugs used (t = 3.51,p<.001), the number of drinks typically consumed per occasion (t = 5.96,p<.001), and the frequency of intoxication in a typical month (t = 5.61,p<.001) (see table 4). Furthermore, participants identified by the SASSI-3 as being at high risk of having a substance use disorder reported feeling significantly more depressed and suicidal than others (t’s = 3.88 and 3.35 respectively, p’s ¸.001, see table 4). Conversely, classifications on the SASSi-3 were found to be disassociated with a number of theoretically unrelated variables such as scores on intelligence, reading, verbal comprehension and arithmetic tests (see table 4).

Conclusion
These results indicate that the SASSI-3 is a reliable and valid measurement tool and support its use for clinical assessment. The SASSI-3 demonstrated acceptable internal consistency and test-retest reliability, and was found to correspond highly with independent clinical diagnoses. The SASSI-3 was also associated with theoretically related criteria (e.g., substance-related arrests and the number of illicit drugs used) but disassociated with theoretically unrelated criteria (e.g., intelligence, reading and arithmetic tests).

The S.A.S.S.I. Institute

SASSI-3 Accuracy In Identifying Use Disorders As A Function Of Ethnic Group

Note: All figures are percentages.

• Prevalence = Rate of clinically diagnosed substance use disorders in the given sample;
• Sensitivity = Percentage of those diagnosed with a substance use disorder who tested positive on the SASSI-3;
• Specificity = Percentage of those diagnosed as not having a substance use disorder who tested negative on the SASSI-3;
• PPP = Positive Predictive Power = Percentage of those who tested positive, who also were diagnosed with a substance use disorder;
• NPP = Negative Predictive Power = Percentage of those who tested negative, who were also diagnosed as not having a substance use disorder;
• Overall Accuracy = Rate of correctly classifying both those who were and those who were not diagnosed with a substance use disorder.

A logistic regression analysis revealed that the racial/ethnic group did not have a statistically significant impact on the accuracy of the SASSI-3. (Improvement X2 = .13,p = .99)

Table 1
Alpha Coefficients for the SASSI-3 and its Subscales

Table 2
Test-Retest Coefficients for the Subscales of the SASSI-3*

FVA = Face Valid Alcohol; FVOD = Face Valid Other Drugs; SYM = Symptoms; OAT = Obvious Attributes
SAT = Subtle Attributes; DEF = Defensiveness; SAM = Supplemental Addiction Measure
FAM = Family Versus Controls; COR = Correctional

Table 3
The Correspondence of Classifications on the SASSI-3 to Clinical
Diagnosis of the Presence of a Substance Use Disorder

Table 4
T-Tests: Means on Theoretically Related and Unrelated Criteria
by Classification Status on the SASSI-3

The SASSI is professionally scored by Proscreening.net

© 1999 Professional Alternatives, Inc.